Austrian business cycle theory (ABCT) explains the boom-bust cycle as a problem of intertemporal discoordination. When monetary policy pushes the market rate of interest below the natural rate, entrepreneurs receive a false signal that real savings are more abundant than they actually are. Long-duration and capital-intensive projects, therefore, appear profitable even though the real resources needed to complete and sustain them have not increased. Ludwig von Mises described this with the analogy of builders acting as if they have more bricks than they actually possess; the boom conceals the shortage, while the bust reveals it (Mises [1912] 1953; [1949] 1998, 557; Garrison 2001).
Real estate development is a natural setting in which to test this logic. Large multifamily projects require substantial upfront capital, long construction periods, complex financing, and expectations about future demand, absorption, and exit pricing. Moreover, they are also highly sensitive to capitalization rates, debt costs, land values, and investor willingness to accept risk. The skyscraper curse literature applies ABCT to the most visible form of this phenomenon: record-setting towers that cluster near the end of credit booms (Lawrence 1999; Thornton 2005; 2018). The key claim is not that skyscrapers cause recessions; rather, it is that exceptionally ambitious projects are symptoms of distorted credit conditions.
This article extends that insight to five-star US multifamily development. In this sector, malinvestment may not appear as record building height. Zoning, site constraints, and market-specific demand can limit physical scale (Glaeser and Gyourko 2018, 5–7). The more informative signal may instead be financial: the compression of the risk premium between construction start and delivery. The research question is therefore whether artificially low rates at construction start correspond to larger projects or to lower compensation for project-level real estate risk upon delivery.
Theory and Prior Literature
The ABCT mechanism begins with the interest-rate signal. In a Wicksellian framework, the natural rate coordinates saving and investment, while the market rate is the observed financing rate. When credit expansion holds the market rate below the natural rate, investment rises in ways not supported by voluntary saving. Friedrich Hayek (2008, 260–61) and Roger Garrison (2001, chap. 4) emphasize that this distortion lengthens the structure of production and increases exposure to later correction when credit conditions normalize. Anthony Carilli and Gregory Dempster (2008) and Levi Russell and Michael Langemeier (2015) operationalize this idea by examining gaps between market rates and proxies for natural rates.
This lag between the initial investment decision and later market repricing is especially important for real estate because development is not a short-horizon trade. A project can be conceived, capitalized, entitled, built, and delivered across very different monetary environments. The developer makes the initial decision under one set of interest rates and asset-price expectations, but the asset is valued under another set at delivery. That lag creates an empirical opportunity. If ABCT is right, the error does not need to be obvious when the project breaks ground. Rather, it should be revealed when the financing environment changes and the spread required to compensate investors for real estate risk is reassessed.
The skyscraper curse literature gives ABCT an expression in real estate. Thornton (2005; 2018) argues that cheap credit affects land prices, technology adoption, and the feasibility of large capital projects. Rising urban land values and falling hurdle rates can make taller buildings appear justified. Empirical work has been mixed. Barr et al. (2015) question whether skyscraper height reliably predicts downturns, while Boyle et al. (2016) defend the curse as a symptom of credit excess rather than a mechanical forecasting rule. This distinction matters: The causal force is not height itself, but the distorted credit environment that allows ambitious projects to proceed.
Data and Method
The underlying research compiled a project-level dataset of US five-star multifamily developments from Q1 2000 to Q1 2025 (Maichel 2025). The sample is restricted to five-star multifamily projects with more than 375 units. The full dataset includes 264 projects; 250 projects are used in the regression models after excluding observations with missing land values and excluding 2020 and 2021 to avoid pandemic-specific shocks. Projects included in the analysis began construction between March 2006 and February 2024, with deliveries occurring between 2008 and 2025. CoStar (1982–2026) provides project characteristics, including rentable building area (RBA), unit count, stories, delivery timing, land values, capitalization rates, market vacancy, and rent growth. Federal Reserve Economic Data, Bloomberg, and Wall Street Journal data provide the monetary and Treasury-rate inputs used to construct the interest-rate gap and risk-premium measures (BEA 1959–2026a; 1959–2026b; Board of Governors of the Federal Reserve System 2016–2021; 2021–2026; Bloomberg 1954–2026; Tullett Prebon 2006–2026).
The core explanatory variable, gap_start in table 1, is the interest-rate gap at the start of construction, defined as the market rate minus the estimated natural rate. Following prior empirical work, the natural-rate proxy is based on the consumption-saving relationship; the market rate is represented by the relevant policy or administered rate series, including the effective federal funds rate and reserve-balance rates. A negative gap_start indicates that the market rate at construction start is below the estimated natural rate and therefore represents the monetary condition most consistent with ABCT.
Two groups of ordinary least squares models are estimated. The first tests physical scale. Separate dependent variables are RBA, unit count, and stories, allowing the analysis to distinguish square footage, unit production, and vertical intensity rather than treating height as the only expression of scale. The second tests financial repricing. The dependent variable is rp_delta, the change in the risk premium from construction start to delivery, where the risk premium equals the market capitalization rate minus the ten-year Treasury yield. A negative rp_delta indicates compression of the risk premium, meaning the project delivers with lower compensation over the risk-free rate than was anticipated at the start of construction. Controls include initial capitalization rate, average land cost per acre, market vacancy, trailing-twelve-month rent growth, and US Census regional fixed effects.
This design intentionally separates a physical-channel test from a pricing-channel test. The physical-channel test asks whether low-rate conditions lead developers to consume more visible real resources—to build more square feet, more units, or more height. The pricing-channel test asks whether low-rate conditions cause investors to accept inadequate risk compensation. The distinction is necessary because a project can be financially malinvested even if its floor area and height look ordinary for the local market.
Results
The descriptive statistics confirm that the sample consists of large, capital-intensive projects. The average project contains roughly 646,500 square feet of RBA and about 530 units, while the average height is approximately 27 stories (Maichel 2025, 47). The average starting interest-rate gap is about −5 percentage points, indicating that many projects began in monetary conditions consistent with artificially easy credit. The average rp_delta is close to zero, but it varies meaningfully across projects, which allows the analysis to test whether monetary conditions explain repricing rather than assuming that all projects experienced the same change in risk premiums between construction start and delivery. The sample also contains substantial dispersion in land costs, vacancy, and rent growth, making it important to control for local market fundamentals rather than attributing all variation to national credit conditions.
The physical scale models do not support the claim that negative interest-rate gaps produce larger five-star multifamily projects. Table 1 shows that the regression coefficient on gap_start is statistically insignificant in the RBA model (p = 0.135), the unit-count model (p = 0.313), and the stories model (p = 0.624). The signs are also negative across the three specifications, which is inconsistent with the simple version of the skyscraper curse prediction. In this sample, regional fixed effects matter more than monetary gaps. Middle Atlantic projects are larger in RBA and units, and some regions differ significantly in height, suggesting that zoning, density, urban form, and local market conditions shape physical scale more than the initial interest-rate gap.
The risk-premium results are stronger. In the gap_start specification, the coefficient on gap_start is 0.156 and is statistically significant at the p < 0.001 level. Because the gap is measured as market rate minus natural rate, a more negative value implies easier credit. The positive coefficient means that deeper negative gaps are associated with a more negative rp_delta, or greater risk-premium compression by delivery. In practical terms, the regression demonstration shows that a positive starting gap of +2.7 percent predicts a widening of about 125 basis points, an average gap of −5.1 percent predicts a widening of roughly 3 basis points, and a deeply negative gap of −12.3 percent predicts a compression of about 110 basis points.
A second risk-premium specification replaces gap_start with gap_delta, the change in the interest-rate gap during construction. The coefficient is −0.0549 and is statistically significant at p < 0.001. Since a negative gap_delta indicates that the gap narrowed during construction, the result suggests that credit normalization is associated with further risk-premium compression. This matters because it shows when the error becomes visible. The initial distortion matters, but so does the movement toward normalization during the construction period. The project may look sustainable when started, but delivery occurs after financing and valuation assumptions have shifted.
Discussion
These findings do not reject the Austrian interpretation; they narrow where we should look for it. The evidence does not show that cheap credit made five-star multifamily projects physically larger or taller. That result is unsurprising once the sector is examined more closely. The scale of multifamily construction is constrained by parcel size, municipal approvals, zoning, neighborhood opposition, unit mix, and local demand. A developer may face distorted capital market signals yet still be unable to add height or units. Physical overbuilding is, therefore, an incomplete indicator of malinvestment. It may also be a lagging or locally filtered indicator. By the time a project appears in the skyline, many of the most important economic decisions have already occurred in the land purchase, debt sizing, return target, capitalization-rate assumption, and equity underwriting.
The financial evidence is more consistent with ABCT. Projects that begin under deeper negative interest-rate gaps experience greater compression in the spread between the capitalization rate and the ten-year Treasury yield. That compression represents diminished compensation for commercial real estate risk. In Austrian terms, the market initially behaves as if more real resources and future demand are available than actually exist. The correction does not necessarily show up as an unfinished tower or a visibly excessive building. It can show up as a capital structure and valuation problem: The project delivers into a world where risk was not priced adequately.
This interpretation also refines the skyscraper curse. The curse is often discussed visually, through the tallest buildings. The present evidence suggests that in institutional multifamily markets, the more meaningful symptom may be compressed risk premiums. The Austrian insight remains intact: Artificially low rates encourage entrepreneurial error in long-duration projects. But the empirical location of that error shifts from architectural scale to financial pricing. The skyscraper is a monument to the boom; the compressed risk premium is the accounting of the same distortion inside the project economics. This also helps reconcile the mixed empirical literature. If researchers look only for height records, they may miss more ordinary projects that still embody boom-era mispricing. A broader ABCT test should therefore ask where the distortion is most likely to be recorded: in the skyline, in the capital stack, or in the spread investors accept for bearing illiquid real estate risk.
Implications and Limitations
The article’s main contribution is to show how malinvestment can be tested at the project level. It compares two possible channels: physical scale and financial repricing. The results suggest that in modern five-star multifamily development, ABCT may be more clearly exhibited through spreads, capitalization rates, and delivery-period repricing than through visible construction excess alone. Practically, this implies that low-rate environments should not be treated merely as periods of cheaper capital; they may also be periods in which risk is systematically underpriced, especially for long-gestation projects exposed to monetary normalization before completion and subsequent stabilization.
Notably, this research has several limitations. The sample is restricted to US five-star multifamily projects and may not generalize to office, retail, industrial, or lower-quality apartment assets. The natural rate of interest is unobservable, so the interest-rate gap depends on a proxy. The models also cannot fully capture developer quality, capital stack structure, hedging, debt terms, entitlement risk, or local political constraints. Finally, delivery is measured at the certificate of occupancy, while additional repricing may occur during lease-up and stabilization.
Future research should extend the test across property types and monetary regimes. Office and hotel development may be more sensitive to employment cycles and business travel, while industrial assets may respond more directly to logistics demand and e-commerce growth. Cross-country comparisons would also be useful because central bank frameworks, land-use systems, and construction finance practices differ materially. Finally, loan-level data would strengthen the analysis by distinguishing projects with fixed-rate debt, floating-rate debt, rate caps, mezzanine financing, preferred equity, and other features that change exposure to monetary normalization. Such data would allow researchers to identify whether the compression documented here is borne primarily by developers, lenders, or equity investors.
Conclusion
The central finding is that malinvestment in five-star multifamily development appears more financial than physical. The evidence does not show that negative interest-rate gaps lead developers to build larger or taller projects. It does, however, suggest that projects started during deeper negative interest-rate gaps experience statistically significant compression in risk premiums by delivery. This matters for ABCT because it identifies a modern channel through which artificially low rates can distort entrepreneurial calculation even when visible overbuilding is limited. It also explains why market participants may underestimate risk. A project may look institutionally acceptable, appropriately located, and of ordinary size, yet remain vulnerable because the price of risk embedded in the deal was set during abnormal monetary conditions.
These findings suggest a broader interpretation of the skyscraper curse. In some markets, the curse may appear as record height. In five-star multifamily, however, this evidence suggests that it appears as a compressed form of compensation for risk in long-duration real estate projects. Mises’s brick analogy still applies, but the missing bricks are financial rather than architectural. Developers and investors do not necessarily run out of physical material; they run out of spread, margin, and properly priced risk when credit conditions move back toward normal.